The invention relates to a vehicle model identification technology, in particular to a hybrid neural network vehicle model identification method based on audio feature fusion. According to the invention, the problem of low recognition accuracy when a traditional vehicle model recognition technology is applied to complex illumination and weather environments is solved. The invention discloses the hybrid neural network vehicle model identification method based on audio feature fusion. The method is implemented by adopting the following steps: step 1, randomly destroying audio signals of a to-be-trained vehicle model: randomly selecting 80% of vehicle audio signals from the destroyed vehicle audio signals, and then superposing environmental noise on the selected vehicle audio signals; step2,constructing the hybrid neural network; step3, inputting the fusion features with the labels into the hybrid neural network for supervised training; and step 4, inputting the vehicle audio signal of the vehicle model to be identified into the trained hybrid neural network. The method of the invention is suitable for vehicle model identification.
本发明涉及车型识别技术,具体是一种基于音频特征融合的杂交神经网络车型识别方法。本发明解决了传统的车型识别技术在应用于复杂的照明和天气环境下时识别准确率低的问题。一种基于音频特征融合的杂交神经网络车型识别方法,该方法是采用如下步骤实现的:步骤一:对待训练车型的音频信号进行随机破坏:从破坏后的车辆音频信号中随机选取80%的车辆音频信号,然后在选取的车辆音频信号上叠加环境噪声;步骤二:构建杂交神经网络;步骤三:将带有标签的融合特征输入到杂交神经网络中进行有监督训练;步骤四:将待识别车型的车辆音频信号输入到训练好的杂交神经网络中。本发明适用于车型识别。
Hybrid neural network vehicle model identification method based on audio feature fusion
一种基于音频特征融合的杂交神经网络车型识别方法
2020-04-17
Patent
Elektronische Ressource
Chinesisch
Fusion Model of Vehicle Positioning with BP Neural Network
British Library Conference Proceedings | 2003
|